The Journal of Physiology

J Physiol 593.8 (2015) pp 1841 1856 1841 Effects of lung ventilation perfusion and muscle metabolism perfusion heterogeneities on maximal O 2 transport and utilization I. Cano 1,2,J.Roca 1,2 and P. D. Wagner 3 1 Hospital Clinic, IDIBAPS, CIBERES, Universitat de Barcelona, Barcelona, Catalunya, Spain 2 Centro de Investigación en Red de Enfermedades Respiratorias (CibeRes), Palma de Mallorca, Spain 3 School of Medicine University of California, San Diego, San Diego, CA 92093-0623A, USA The Journal of Physiology Key points We expanded a prior model of whole-body O2 transport and utilization based on diffusive O 2 exchange in the lungs and tissues to additionally allow for both lung ventilation perfusion and tissue metabolism perfusion heterogeneities, in order to estimate V O2 and mitochondrial P O2 (P mo2 ) during maximal exercise. Simulations were performed using data from (a) healthy fit subjects exercising at sea level and at altitudes up to the equivalent of Mount Everest and (b) patients with mild and severe chronic obstructive pulmonary disease (COPD) exercising at sea level. Heterogeneity in skeletal muscle may affect maximal O2 availability more than heterogeneity in lung, especially if mitochondrial metabolic capacity ( V MAX ) is only slightly higher than the potential to deliver O 2,butwhen V MAX is substantially higher than O 2 delivery, the effect of muscle heterogeneity is comparable to that of lung heterogeneity. Skeletal muscle heterogeneity may result in a wide range of potential mitochondrial P O2 values, a range that becomes narrower as V MAX increases; in regions with a low ratio of metabolic capacity to blood flow, P mo2 can exceed that of mixed muscle venous blood. The combined effects of lung and peripheral heterogeneities on the resistance to O2 flow in health decreases with altitude. Abstract Previous models of O 2 transport and utilization in health considered diffusive exchange of O 2 in lung and muscle, but, reasonably, neglected functional heterogeneities in these tissues. However, in disease, disregarding such heterogeneities would not be justified. Here, pulmonary ventilation perfusion and skeletal muscle metabolism perfusion mismatching were added to a prior model of only diffusive exchange. Previously ignored O 2 exchange in non-exercising tissues was also included. We simulated maximal exercise in (a) healthy subjects at sea level and altitude, and (b) COPD patients at sea level, to assess the separate and combined effects of pulmonary and peripheral functional heterogeneities on overall muscle O 2 uptake ( V O2 ) and on mitochondrial P O2 (P mo2 ). In healthy subjects at maximal exercise, the combined effects of pulmonary and peripheral heterogeneities reduced arterial P O2 (P ao2 ) at sea level by 32 mmhg, but muscle V O2 by only 122 ml min 1 ( 3.5%). At the altitude of Mt Everest, lung and tissue heterogeneity together reduced P ao2 by less than 1 mmhg and V O2 by 32 ml min 1 ( 2.4%). Skeletal muscle heterogeneity led to a wide range of potential P mo2 among muscle regions, a range that becomes narrower as V MAX increases, and in regions with a low ratio of metabolic capacity to blood flow, P mo2 can exceed that of mixed muscle venous blood. For patients with severe COPD, peak V O2 was insensitive to substantial changes in the mitochondrial characteristics for O 2 consumption or the extent of muscle heterogeneity. This integrative computational model of O 2 transport and utilization offers the potential for estimating profiles of P mo2 both in health and in diseases DOI: 10.1113/jphysiol.2014.286492

1842 I. Cano and others J Physiol 593.8 such as COPD if the extent for both lung ventilation perfusion and tissue metabolism perfusion heterogeneity is known. (Resubmitted 29 October 2014; accepted after revision 27 January 2015; first published online 26 February 2015) Corresponding author I. Cano: IDIBAPS, C/Villarroel 170, 08036, Barcelona, Spain. Email: iscano@clinic.ub.es Abbreviations COPD, chronic obstructive pulmonary disease; DL, lung diffusing capacity; DM, tissue (muscle) diffusing capacity; [Hb], haemoglobin concentration; FEV, forced expired volume; F I,O2,inspiredO 2 fraction; P ao2, arterial P O2 ; P mo2, mitochondrial P O2 ; Q, blood flow distribution; V A, expired ventilation; V A / Q, ventilation perfusion inequality; V I, inspired ventilation; V O2max, maximum oxygen delivery uptake; V max, muscle metabolism; V MAX, mitochondrial maximal capacity for O 2 consumption. Introduction Current modelling of the O 2 pathway (Wagner, 1993, 1996; Cano et al. 2013) is based on the concept that it is the functional integration amongst all individual components of the oxygen transport and utilization system (i.e. lungs and chest wall, heart, blood and circulation, and tissue mitochondria) which determines maximal overall O 2 uptake ( V O2 max). In these models, important simplifying assumptions were made. In particular, O 2 exchange within the lungs was simplified by ignoring ventilation perfusion ( V A / Q) inequality, and exchange focused only on alveolar capillary diffusion. This was felt to be reasonable as the minimal V A / Q inequality found in health (Wagner et al. 1974) is normally of minor importance to O 2 exchange, while allowing for such inequality greatly increased model complexity. However, this would be an unacceptable simplification in cardiopulmonary diseases where V A / Q inequality can be substantial. Similarly, within the muscles, potential heterogeneity of local metabolic capacity and demand ( V O2 )inrelationtobloodflowdistribution( Q) wasnot considered, and while diffusive movement of O 2 from the muscle microcirculation to the mitochondria was modelled, the muscles were considered to be functionally homogeneous, i.e. with blood flow perfectly matched to O 2 demand throughout. Much less information exists about normal V O2 / Q heterogeneity in muscle (Richardson et al. 2001), but using near infrared spectroscopy (NIRS), preliminary data in normal subjects (Vogiatzis et al. 2015) suggest that V O2 / Q matching is much tighter than is V A / Q matching. Thus, V O2 / Q dispersion in muscle appears to be only about ¼ of V A / Q dispersion in normal lung (Wagner et al. 1987a,b). Another simplifying assumption was that all of the cardiac output flowed to the exercising muscles, which is clearly untrue. This is not of great quantitative importance to muscle O 2 transport in health because of the ability to increase cardiac output to over 20 l min 1,renderingthe fraction of cardiac output perfusing non-exercising tissue small. However, such an assumption would pose a substantial limitation if the models were applied to patients with chronic cardiorespiratory disease where peak exercise may elicit less than a doubling of the resting cardiac output and resting V O2. The purpose of the study reported in the current paper was therefore to first expand the prior O 2 transport pathway model (Wagner, 1993, 1996; Cano et al. 2013) by allowing for (a) V A / Q heterogeneity in the lung, (b) V O2 / Q heterogeneity in the muscle, and (c) perfusion and metabolism in non-exercising tissues, and then to analyse the impact of possible heterogeneity in lungs and in peripheral exercising tissues, separately and combined, on overall O 2 transport and utilization at maximum exercise, both in health and in disease. In health, we simulated exercise not just at sea level but at altitudes up to the summit of Mt Everest using as input data variables obtained in Operation Everest II (Sutton et al. 1988). In disease, we chose COPD as the example of pulmonary gas exchange heterogeneity because of availability of O 2 transport data (Blanco et al. 2010), and because of general interest in muscle function in this disease. We then posed four questions: (a) in health, what are the potential effects of lung and muscle heterogeneity (each modelled over a wide range) on oxygen flux and partial pressuresatallstepsintheo 2 pathway at sea level; (b) what is the impact of typical normal levels of heterogeneity on these variables when O 2 availability is reduced by exposure to high altitude; (c) how substantial variations of the three least well established determinants of O 2 flux and partial pressures in the transport utilization pathway ( V MAX and P 50 of the mitochondrial respiration curve and the extent of muscle heterogeneity) would affect both V O2 and mitochondrial P O2 estimates; and (d) what is the effect of (measured) lung and possible muscle heterogeneity on overall oxygen uptake and mitochondrial P O2 in mild and severe COPD, where O 2 transportislessthannormal. Methods The prior O 2 transport and utilization model In the absence of heterogeneity in lungs or tissues, and ignoring metabolism and blood flow to non-exercising tissues, the system describing maximal O 2 transport and

J Physiol 593.8 utilization (Wagner, 1993, 1996; Cano et al. 2013) was composed of five mass conservation equations as shown in Fig. 1, which is reproduced from Cano et al. (2013) for the convenience of the reader. These equations represent: (1) O 2 transport by ventilation from the atmosphere to alveolar gas (eqn (1) in Fig. 1); (2) O 2 diffusion from the alveolar gas to the lung capillaries (eqn (2) in Fig. 1); (3) O 2 transport through the systemic circulation (eqn (3) in Fig. 1); (4) O 2 diffusion from systemic capillary blood to the muscle mitochondria (eqn (4) in Fig. 1); and (5) mitochondrial O 2 utilization (eqn (5) in Fig. 1). The inputs to these equations are, at maximal exercise: inspired O 2 fraction (F I,O2 ), ventilation ( V I, inspired; V A, expired), lung diffusing capacity (DL), cardiac output ( Q), acid base status, haemoglobin concentration ([Hb]), tissue (muscle) diffusing capacity (DM) and the characteristics of mitochondrial respiration (modelled as a hyperbolic curve with its two parameters, V MAX and P 50, as defined below and illustrated in Fig. 2). The five outputs, or unknowns, in these equations are the maximal rate of O 2 uptake ( V O2 max) and alveolar, systemic arterial, venous and mitochondrial P O2 values. While the focus is clearly on O 2 transport, CO 2 cannot be ignored in such a model because of the effects of CO 2 retention, for example on the O 2 Hb dissociation curve and on alveolar P O2. Thus, the same processes and Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1843 equations are used for CO 2, (West, 1971), treating CO 2 at every step side by side with O 2 using the published CO 2 dissociation curves and acid base relationships based on Kelman s subroutines (Kelman, 1966) such that O 2 CO 2 interactions are accounted for both in the lungs and muscle. Furthermore, the effect of alveolar ventilation on arterial P CO2 is also innately included as one of the input variables needed is total ventilation. An exception is made for mitochondrial respiration, where, because the mitochondrial respiration curve for O 2 consumption does not apply to CO 2,CO 2 production is related to V O2 through a stipulated value of the respiratory quotient, and the solutions for both gases sought simultaneously. Modelling lung ventilation perfusion inequality To expand the model and allow for ventilation perfusion ( V A / Q) inequality, ventilation and perfusion within the lung were distributed amongst a set of virtual lung compartments, each defined by its own ratio of ventilation ( V A ) to blood flow ( Q), with the compartments connected in parallel. This is exactly the same approach as first advanced by West in 1969 (West, 1969). Then, the computation for diffusive alveolar capillary exchange, previously applied to just a single lung compartment representing the homogeneous lung, is now applied to all Figure 1. Schematic representation of the oxygen transport and utilization system The schematic representation is shown with the five associated mass conservation equations governing O 2 transport (eqns (1) (4)) and utilization (eqn (5)). Reproduced from Cano et al. (2013) with permission.

1844 I. Cano and others J Physiol 593.8 compartments in turn. The entering mixed venous blood is the same for all compartments, and the task is to compute the P O2 and P CO2 values expected at the end of the capillary in each compartment of different V A / Q ratio. This is done with a forward numerical integration procedure (Wagner et al. 1974), also used in Cano et al. (2013), and the outcome for each compartment depends on the compartmental DL/ Q and V A / Q ratios, the inspired gas and mixed venous blood O 2 and CO 2 composition, and the shape and position of the respective binding curves in blood. This approach is in fact the same as developed by Hammond & Hempleman (1987), and which is also contained within the multiple inert gas elimination technique (Wagner, 2008). Once all compartments have been subjected to this numerical integration, O 2 and CO 2 concentrations in their effluent blood are averaged in a perfusion-weighted manner to compute the O 2 and CO 2 profile of the systemic arterial blood that will reach both the exercising muscles and non-exercising tissues and organs. In the present application, two means of data entry are possible for specifying the features of the V A / Q distribution, allowing flexibility. Individual compartmental values for ventilation and blood flow can be used, or the compartments can be calculated based on a V A / Q distribution containing one or more modes, using for each mode its first three moments (mean, dispersion and skewness) as originally programmed by West (1969). Both types of inputs end up providing a multi-compartmental V A / Q distribution (indicated in Fig. 1). For simplicity, when computing the lung compartments based on a V A / Q distribution, the lung diffusing capacity (DL) was always distributed in proportion to compartmental blood flow, so that compartmental DL/ Q remains constant throughout Figure 2. Graphical depiction of the hyperbolic equation for oxidative phosphorylation The graphical depiction of the equation for oxidative phosphorylation (eqn (1) in Fig. 1) is fitted to data of Scandurra & Gnaiger (2010; p. 16, Fig 3B). the lung. This decision does not eliminate the opportunity to introduce actual data on DL/ Q distribution when available. In summary, to allow for V A / Q heterogeneity, the single homogeneous lung compartment in the prior model has been replaced by a multi-compartment model built from the algorithms of West (1969) as modified by Hammond & Hempleman (1987) and which allows for both V A / Q inequality and diffusive exchange. Just as in the prior model, necessary inputs remain the composition of inspired gas and mixed venous blood, and the output remains the composition of systemic arterial blood destined for the muscles and other tissues. Modelling peripheral heterogeneity Similarly to lung ventilation perfusion mismatch, tissue heterogeneities are considered as (regional) variability in a functional ratio. However, instead of V A / Q being that ratio as in the lung, it is the ratio of mitochondrial metabolic capacity ( V MAX )tobloodflow( Q) that is considered in the muscles. Here, V MAX is the local mitochondrial maximal metabolic capacity to use O 2, and is the very same parameter as V MAX in eqn (5), Fig. 1. The ratio V MAX / Q in essence expresses the balance between O 2 supply ( Q) and O 2 demand ( V MAX ). In the prior model, the muscle was considered homogeneous in terms of V MAX / Q ratios, but now the muscle is considered as a parallel collection of muscle units each with its own ratio of V MAX / Q.Exactly as with the lung, the V MAX / Q distribution can be specified either as a set of individual compartmental values for Q, DM, V MAX and P 50, or computed from the moments of a(multimodal) V MAX / Q distribution. While it would be possible to assign every V MAX / Q compartment a unique mitochondrial P 50 (see eqn (5) in Fig. 1), this has not been modelled to date, and P 50 has been taken to be constant throughout any single muscle. Also as in the lung, diffusion is considered in every V MAX / Q unit, with the diffusing capacity DM distributed in proportion to Q, just as for the lung. However, the algorithm can accommodate different values across compartments for both P 50 and DM if desired. The same forward integration procedure is used as in the prior model (Cano et al. 2013), when it was applied to the entire muscle considered as one homogeneous unit to compute each muscle compartment s effluent venous O 2 and CO 2 levels for the given inflowing arterial blood composition, and the V MAX / Q and DM/ Q ratios of the unit. Again as in the lung, the effluent venous blood from all muscle units is combined, or mixed, in terms of both O 2 and CO 2 concentrations, and a resulting mixed muscle venous P O2 and P CO2 is then calculated. Finally, this blood is mixed with that from the non-exercising tissues (see

J Physiol 593.8 below). This mixed venous blood returns to the lungs for re-oxygenation. In summary, to allow for V MAX / Q heterogeneity, the single homogeneous muscle compartment in the prior model has been replaced by a multi-compartment model which allows for both V MAX / Q inequality and diffusion impairment. Similarly to the prior model, necessary additional inputs remain the composition of arterial blood, and the blood binding characteristics of O 2 and CO 2. The output remains the composition of mixed muscle venous blood destined for return to the lungs for re-oxygenation. In this section, the expression of peripheral heterogeneity using the oxygen flow ( Q O2 ) instead of blood flow ( Q) was considered, but since at any given arterial P O2, the arterial blood entering any muscle region has the same oxygenation, the distribution of V MAX / Q will not differ in dispersion from that of V MAX / Q delivery. Moreover, O 2 concentration is an outcome variable. Hence, we cannot in practice adopt O 2 delivery as the denominator. Modelling metabolism and blood flow to non-exercising tissues Non-exercising tissue blood flow and metabolism are incorporated simply by assigning values for total non-exercising tissue V O2 and Q, and using the Fick principle to compute the tissue venous O 2 concentration. Then, a mixing equation combines venous blood from this non-exercising tissue with that from the mixed muscle venous blood (see above) in proportion to the blood flow rates assigned to the non-exercising tissues and muscles respectively, to form what will be pulmonary arterial blood reaching the lungs for re-oxygenation. Achieving a solution for the model In the prior model it was explained (Cano et al. 2013) how the model was run and a solution achieved. The same process is used with the above-described expansions in the current paper. In brief, a starting estimate is made for the composition of mixed venous blood entering the lungs, and for that estimate, the lung component is executed over the many V A / Q compartments and the end-capillary O 2 and CO 2 concentrations averaged over all compartments (weighted by compartmental blood flow) to yield an estimate of systemic arterial P O2 and P CO2. These are then used as input data for the muscle part of the system, and that component is run once for each V MAX / Q compartment. The effluent compartmental venous O 2 and CO 2 levels are then averaged (again blood flow-weighted), and mixed with the blood draining the non-exercising tissues (see previous section). It is the P O2 and P CO2 of this Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1845 mixed blood that is the new estimate at mixed venous P O2 and P CO2 for the next iteration (or cycle) of pulmonary gas exchange: To this point, all that has changed between the first and second cycles is the composition of mixed venous blood. After pulmonary gas exchange calculations using this new venous blood composition are complete, new values of arterial blood P O2 and P CO2 are calculated, and now muscle O 2 exchange is recomputed, from which new values of mixed venous P O2 and P CO2 are calculated. This cycling back and forth between the lungs and the tissues is continued until the five V O2 estimates one computed from each of the five equations in Fig. 1 are identical (to within ± 0.1mlmin 1 ). At this point, stability of all P O2 values (alveolar, arterial, muscle venous and mitochondrial) in the face of further cycling between the lungs and tissues is also achieved. When this is observed, the governing requirement for conservation of mass has been achieved, and this signals the end of the run. Analysis and input data for simulations As mentioned above, four questions were posed, and these are discussed below. First, in health, we assessed the effects of a wide range of lung and tissue functional heterogeneities on oxygen tensions and utilization at all steps of the O 2 pathway. This extensive exploration was carried out for conditions of maximal exercise breathing room air at sea level. To this end, input data defining O 2 transport conductances (i.e. ventilation, cardiac output, lung and muscle diffusional conductances) from normal subjects exercising maximally at sea level in Operation Everest II (Sutton et al. 1988) were used just as for the previous models (Wagner, 1993, 1996). Lung heterogeneity was analysed from complete homogeneity (logsd Q = 0, second moment of the blood flow ( Q) distribution on a logarithmic V A / Q scale equals zero), to very high inhomogeneity (logsd Q = 2) as might be seen in the critically ill. We also explored the effects of skeletal muscle heterogeneity on overall O 2 uptake ( V O2 max) using a similar wide range of logsd V max (second moment of the muscle metabolism ( V max ) distribution on a logarithmic V max / Q scale equals zero) from 0 to 2. The effects of lung and skeletal muscle heterogeneity on oxygen transport and utilization are displayed in Figs 3 and 4. In Fig. 3, and the first section in Results, the impact of skeletal muscle heterogeneity on V O2 max is assessed at two different levels of V max corresponding to 120% and 250% V O2 max. The rationale behind a value for V max of4.6lmin 1, which is 20% higher than the measured V O2 max (3.8 l min 1 ) at sea level in Operation Everest II (Sutton et al. 1988), is that it is a conservative estimate. This is so, because it is known that measured V O2 max is less than mitochondrial capacity to use O 2 (i.e. V max ) in healthy

1846 I. Cano and others J Physiol 593.8 fit subjects and exercise capacity is known to be increased within this range when breathing 100% O 2 (Welch, 1982; Knight et al. 1992). However, we acknowledge that reports on subjects exercising a limited amount of muscle mass, (i.e. knee extensor of one limb); Andersen & Saltin, 1985) indicate the potential for markedly higher V max values up to 250% of V O2 max (which would scale to 9.5 l min 1 for V O2 max measured in subjects from Operation Everest II). In the study by Andersen & Saltin, a rate of 350 ml kg 1 was measured, which extrapolated to a conservative estimate of 27 kg of exercising muscle mass corresponds to a V O2 max valueof9.5lmin 1. Consequently, these two V max values (4.6 and 9.5 l min 1 ) were used for the simulations displayed in Fig. 3. For the mitochondrialp 50,weusedavalueof0.3mmHg, similar to what has been found experimentally in vitro (Wilson et al. 1977; Gnaiger et al. 1998; Scandurra & Gnaiger, 2010). For non-exercising body tissues we used typical resting total values of V O2 (300 ml min 1 ), V CO2 (240 ml min 1 ) and blood flow (20% of total blood flow at peak exercise). The second question examined the effects of heterogeneity in healthy lungs and muscle on O 2 transport and utilization at altitude. For the lung we used logsd Q = 0.5 as a value commonly seen during exercise (Wagner et al. 1987a,b).Thisvalueisneartheupper end of the normal range, which is 0.3 0.6 (Wagner et al. 1987a,b). In muscle, we used preliminary unpublished estimates of logsd V max = 0.1. Again for P 50, we used a value of 0.3 mmhg and for V max we used a conservative value 20% higher than the measured V O2 max. The input data defining the O 2 transport conductances again came from normal subjects exercising maximally at sea level and altitude in Operation Everest II (Sutton et al. 1988) as used above and again typical resting total values of V O2 (300 ml min 1 ), V CO2 (240 ml min 1 ) and blood flow (20% of total blood flow at peak exercise) were considered for non-exercising body tissues. However, to study the consequences at more altitudes than were examined in Operation Everest II (which were sea level, 4600 m, 6100 m, 7600 m and 8848 m), O 2 transport conductance parameters were linearly interpolated (Table 1) at 305 m (1000 ft) elevation increments from the data obtained at each of the five altitudes studied in Operation Everest II (Sutton et al. 1988). For the third and forth questions we assessed how substantial variations of the three least well established determinants of O 2 flux and partial pressures in the transport utilization pathway ( V MAX and P 50 of the mitochondrial respiration curve andlogsd V MAX ) would affect both V O2 and mitochondrial P O2 estimates, in health (third question) and in COPD (fourth question). Input data defining O 2 transport conductances and lung heterogeneity (i.e.logsd Q = 0.5) from normal subjects exercising maximally at sea level in Operation Everest II Figure 3. Effects of potential lung and muscle heterogeneities on maximal O 2 transport and utilization Independent effects of lung V A / Q heterogeneity (filled circles) and muscle V MAX / Q heterogeneity (open circles) on maximal muscle oxygen uptake ( V O2 max), using input data defining O 2 transport conductances from normal subjects exercising maximally at sea level in Operation Everest II (Sutton et al. 1988). Results shown for V MAX 20% higher than measured V O2 max (continuous line) and V MAX 2.5-fold higher than measured V O2 max (dotted line). Healthy subjects show pulmonary logsd Q between 0.3 and 0.6 (Wagner et al. 1974), whereas skeletal muscle heterogeneity (logsd V MAX ) presents an average value of 0.10 (Vogiatzis et al. 2015). In patients with moderate to severe COPD, logsd Q can present values close to 1.0, but no published data on logsd V MAX are available. Finally, critically ill patients admitted in the Intensive Care Unit (ICU) may show lung heterogeneity values close to 2.0. Again, no information on logsd V MAX is available.

J Physiol 593.8 Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1847 Table 1. Input parameters for the modelling of the oxygen transport system in health at sea level and altitude Barometric Body Haemoglobin Alveolar Blood flow Total lung O 2 Total muscle O 2 Altitude pressure temperature concentration ventilation ( Q; diffusing capacity diffusing capacity (ft) (P B ; mmhg) (T; C) ( [ Hb ] ;gdl 1 ) ( V (BTPS); l min 1 ) l min 1 ) (DL; ml min 1 mmhg 1 ) (DM; ml min 1 mmhg 1 ) 0 760 38 14.2 128.3 25.0 50.9 104.6 5000 639 38 14.9 137.1 23.5 60.9 97.1 10,000 534 38 15.5 145.9 22.0 70.9 89.6 15,000 442 38 16.2 154.6 20.5 80.9 82.1 16,000 426 37.9 16.3 156.4 20.2 82.9 80.6 17,000 410 37.8 16.5 158.1 19.9 84.9 79.1 18,000 394 37.7 16.6 159.9 19.6 86.9 77.6 19,000 380 37.6 16.7 161.6 19.3 88.9 76.1 20,000 365 37.5 16.9 163.4 19.0 90.9 74.6 21,000 351 37.4 17.0 165.2 18.7 92.9 73.1 22,000 337 37.3 17.1 166.9 18.4 94.9 71.6 23,000 324 37.2 17.2 168.7 18.1 96.9 70.1 24,000 311 37.1 17.4 170.4 17.8 98.9 68.6 25,000 299 37 17.5 172.2 17.5 100.9 67.1 26,000 286 37 17.6 173.9 17.2 102.9 65.6 27,000 275 37 17.8 175.7 16.9 104.9 64.1 28,000 264 37 17.9 177.4 16.6 106.9 62.6 29,000 253 37 18.0 179.2 16.3 108.9 61.1 30,000 243 37 18.2 181.0 16.0 110.9 59.6 (Sutton et al. 1988) were used in health. In disease, measured O 2 conductances and lung heterogeneity data came from two previously studied COPD patients (Blanco et al. 2010) exercising maximally (Table 2), one with mild (FEV 1 = 66% predicted post-bronchodilator) and one with severe (FEV 1 = 23% predicted post-bronchodilator) COPD. However, even if reasonable, the values of V MAX and P 50 assumed to answer the previous question are uncertain, let alone whether there is significant V MAX / Q heterogeneity. Because of this, we carried out simulations over a wide range of possible values of these variables, as indicated in Fig. 6. In the case of V max, we used a values of 4.2, 4.6 and 4.9 l min 1 (10, 20% and 30% higher than actual V O2 max, respectively). Moreover, a V max value of9.5lmin 1 (250% V O2 max) was also considered in health because of the rationale described above. For mitochondrial P 50, we used values lower (0.14 mmhg) and higher (0.46 mmhg) than 0.3 mmhg, the value found experimentally in vitro (Wilson et al. 1977; Gnaiger et al. 1998; Scandurra & Gnaiger, 2010). Finally, to assess how sensitive the outcomes are to the degree of heterogeneity, we also carried out calculations with logsd V MAX values of 0.1 (healthy subject estimates), 0.2 and 0.3. With respect to non-exercising body tissues, we used typical resting total values of V O2 (300 ml min 1 ), V CO2 (240 ml min 1 ) and blood flow (20% of total blood flow at peak exercise) for normal subjects. For the two previously studied COPD patients (Blanco et al. 2010) exercising maximally we used measured resting values of whole body V O2 ( V O2 rest and V CO2 (Table 2), along with a blood flow estimate ( Q non exercising )fromthe following formula, which expresses the concept that blood flow is proportional to metabolic rate: Q non exercising = Q max ( V O2 rest/ V O2 max). Results Effects of potential lung and muscle heterogeneities on O 2 transport and utilization in healthy subjects exercising maximally at sea level Figure 3 shows the independent effects of lung V A / Q heterogeneity (filled circles) and muscle V MAX / Q heterogeneity (open circles) on maximal muscle oxygen uptake ( V O2 max), in normal subjects exercising maximally at sea level. In this figure, both V A / Q and V MAX / Q heterogeneity has been varied over a wide range of log SD, from 0 to 2.0. Healthy subjects show pulmonary logsd Q between 0.30 and 0.60 (Wagner et al. 1987a), whereas skeletal muscle heterogeneity (logsd V max / Q) presents an average value of 0.10 (Vogiatzis et al. 2015). In patients with moderate to severe COPD, logsd Q can present values close to 1.0, but no published data on logsd V max / Q are available. Finally, critically ill patients admitted in the Intensive Care Unit (ICU) may show lung heterogeneity values close to 2.0, but no information on logsd V max / Q is available. The main result from Fig. 3 is that heterogeneity in muscle affects O 2 availability

1848 I. Cano and others J Physiol 593.8 Table 2. Input parameters for the modelling of the oxygen transport system in disease (moderate and severe COPD) Moderately limited Severely limited Parameter transport transport Forced expired volume in the 1st second (FEV 1 ; % predicted post-bronchodilator) 66 23 Forced vital capacity (FVC; % predicted post-bronchodilator) 84 33 FEV 1 /FVC 0.58 0.54 Log SD Q 0.67 0.86 Barometric pressure (P B ; mmhg) 758 765 Fractional inspired oxygen (F I,O2 ) 0.2093 0.2093 Haemoglobin concentration ([Hb]; g dl 1 ) 14.4 12.5 O 2 dissociation curve (P 50 ; mmhg) 26.8 26.8 Body temperature (T; C) 37.0 36.8 Weight (kg) 60 36 Body mass index (BMI; kg m 2 ) 19.8 15.2 Resting ventilation ( V (BTPS); l min 1 ) 7.11 7.44 Resting cardiac output ( Q; lmin 1 ) 3.68 3.20 Resting V O2 (ml min 1 ) 235 159 Resting V CO2 (ml min 1 ) 174 103 Resting arterial P O2 (P ao2 ; mmhg) 80 59 Resting arterial P CO2 (P aco2 ; mmhg) 35 46 Resting lactate (mmol l 1 ) 1.47 0.81 Exercise ventilation ( V (BTPS); l min 1 ) 35.3 8.2 Exercise cardiac output ( Q; lmin 1 ) 7.87 3.98 Exercise ( V O2 ;ml min 1 ) 914 355 Exercise arterial P O2 (P ao2 ; mmhg) 90 59 Exercise arterial P CO2 (P aco2 ; mmhg) 31 49 Exercise lung O 2 diffusing capacity (DL; ml min 1 mmhg 1 ) 100 11 Exercise muscle O 2 diffusing capacity (DM; ml min 1 mmhg 1 ) 26 9 Exercise lactate (mmol l 1 ) 3.55 2.03 V MAX more than does heterogeneity in the lungs if is only 20% higher than V O2 max. However, using the higher V MAX (i.e. 9.5 l min 1 ), the effects of lung and muscle heterogeneity are similar. On the other hand, the impact of lung heterogeneity on V O2 max is insensitive to the V MAX values used. In addition, because muscle heterogeneity appears to be much less than that in the lung, both in health and COPD, based on a combination of published and unpublished data as mentioned, actual muscle heterogeneity has less of an impact on O 2 transport than observed lung heterogeneity. The six panels in Fig. 4 display the estimated (combined and separate) effects of lung and muscle heterogeneities on the O 2 transport and utilization system. The upper-left panel indicates arterial P O2 (P ao2 ) over a range of exercising muscle V MAX / Q mismatching (0 to 0.5), expressed as logsd V max. For each logsd V max value, the panel displays P ao2 values for different levels of lung V A / Q heterogeneity, from the homogeneous lung logsd Q = 0) to a highly heterogeneous lung (logsd Q = 2). Each line represents the outcome at a given level of lung heterogeneity over a range of muscle heterogeneity. This panel shows that muscle heterogeneity has only a small effect on P ao2 at all levels of lung functional heterogeneities. However, as is well known, V A / Q inequality has a major impact on arterial oxygenation as shown. Likewise, the upper-right panel shows the effects of muscle heterogeneity on effluent muscle venous P O2 (P vo2 ). The impact is moderate, and is greater than the effect on P ao2. However, it is quantitatively important only at high levels of tissue heterogeneity, especially in combination with lungs containing little heterogeneity. The two middle panels of Fig. 4 show the relationships between mitochondrial P O2 (P mo2 ) among muscle regions (compartments) and skeletal muscle V MAX / Q functional heterogeneity. The left hand panel shows the lowest compartmental values of P mo2 (i.e. when V MAX / Q is high) and the right hand panel shows the highest compartmental values of P mo2 (i.e. when V MAX / Q is low). Note that the ordinate scales are very different in these two panels. These panels show that heterogeneity of V MAX / Q ratios markedly expands the range of P mo2 levels among exercising skeletal muscle compartments. As in the top panels, each line in the two middle panels correspond to different levels of lung heterogeneity from a homogeneous lung (logsd Q = 0) to a highly heterogeneous lung (logsd Q = 2).

J Physiol 593.8 Theleftmiddlepanelshowsthat,asexpected,thegreater the peripheral heterogeneity, the lower the minimum P mo2 values. Moreover, we observe a moderate impact of lung heterogeneity on minimum P mo2, but such impact decreases as skeletal muscle heterogeneity increases. Likewise, the right middle panel shows that skeletal muscle compartments with lowest V MAX / Q ratios may generate Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1849 exceedingly high P mo2 estimates that increase with tissue heterogeneity. Homogeneous lungs show the highest P mo2 values for a given level of peripheral tissue functional heterogeneity. The lower left panel indicates that muscle heterogeneities have small effects on total muscle V O2, especially compared with lung heterogeneities that show Figure 4. Combined and separate effects of lung and muscle heterogeneities on the O 2 transport and utilization system Combined and separated effects of varying degrees of lung ventilation perfusion heterogeneity (SDQ) and varying degrees of exercising muscle mitochondrial metabolic capacity blood flow heterogeneities (logsd V MAX )onthe O 2 transport pathway: P O2 on arterial blood (upper left panel), venous blood (upper right panel) and mitochondria (two central panels), maximal muscle oxygen uptake (lower left panel), and, maximum degree of oxygen unbalance between mitochondrial (P mo2 ) of muscle compartments and mean effluent venous (P vo2 ) oxygen levels (lower-right panel). See text for details.

1850 I. Cano and others J Physiol 593.8 a significant impact on muscle V O2 max. Finally, the lower right panel displays the maximum estimates for P mo2 among muscle regions (as seen in the middle right panel of the same figure) plotted against the corresponding mean effluent muscle P vo2 (from the top right panel). The individual lines indicate different levels of muscle heterogeneities, from a homogeneous tissue (logsd V max / Q = 0) at the bottom to heterogeneous muscle (logsd V max / Q = 0.5) at the top. In the panel, the dashed straight line corresponds to the identity line. The point of presenting this relationship is to show that when the dispersion of muscle V MAX is 0.2 or greater, some regions of muscle will have mitochondrial P O2 values that exceed P O2 of the effluent muscle venous blood. The role of functional heterogeneities, lung and muscle, on O 2 transport and utilization in healthy subjects at altitude The four panels in Fig. 5 display the effects of altitude on four key outcome variables at maximal exercise arterial, muscle venous, and mitochondrial P O2 and V O2 max itself. Two relationships are shown in each panel that computed in the absence of either lung or muscle Figure 5. Role of functional heterogeneities, lung and muscle, on the O 2 transport and utilization system in healthy subjects at altitude Effects of functional inhomogeneity on arterial P O2 (upper left panel), venous P O2 (upper right panel), mitochondrial P O2 (lower left panel) and V O2 (lower right panel) at sea level and altitude. Functional inhomogeneity is expressed as the combined effects of lung ventilation perfusion inequalities and skeletal muscle perfusion metabolism mismatching. Mitochondrial respiration parameters were fixed to a conservative V MAX of 4584 ml min 1 and a P 50 of 0.3 mmhg. Filled circles correspond to simulations outputs when considering both homogeneous lung and tissues. Open circles represent simulation outputs when using reasonable estimates for lung ventilation perfusion (logsd Q of 0.5), tissue perfusion metabolism (logsd V MAX of 0.1), and the metabolic rate of non-exercising tissues (with an assigned V O2 of 300 ml min 1 and a V CO2 of 240 ml min 1 ) and assuming that 20% of total blood flow goes to non-exercising tissues. See text for details.

J Physiol 593.8 heterogeneity, and that computed allowing for reasonable normal estimates of heterogeneity in each location as listed aboveinthemethodssection( V A / Q heterogeneity in the lung quantified by a dispersion value (logsd Q) of 0.5, V MAX / Q heterogeneity in muscle quantified using an estimate of dispersion (logsd V MAX ) of 0.1, non-exercising body tissues using typical resting total values of (300 ml min 1 ), V CO2 (240 ml min 1 ) and 20% of total blood flow at peak exercise, and a conservative V max value 20% higher than the measured V O2 max). Except for arterial P O2 at sea level and altitudes up to about 10,000 ft, Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1851 V O2 the effects of typical levels of heterogeneity are seen to be small, and diminish progressively with increasing altitude. Sensitivity of outcomes to mitochondrial respiration and muscle heterogeneity estimates in health Recall that the three least well established determinants of O 2 flux and partial pressures in the transport utilization pathway are the V MAX and P 50 of the mitochondrial respiration curve (that links V O2 to mitochondrial P O2 ) Figure 6. Sensitivity analysis of mitochondrial respiration and muscle heterogeneity estimates in health Effects of muscle bioenergetics (mitochondrial respiration V MAX and P 50 parameters) and muscle peripheral heterogeneities on muscle V O2 and mitochondrial P O2 (P mo2 ) in health at sea level. V MAX is taken as 10% (upper left panel), 20% (upper right panel), 30% (lower left panel) and 250% (lower right panel) greater than measured V O2 max. For each V MAX, three mitochondrial P 50 values (0.14, 0.3 and 0.46 mmhg) and three degrees (0.1, 0.2 and 0.3) of muscle perfusion metabolism inhomogeneity (logsd V MAX ) were evaluated. See text for details.

1852 I. Cano and others J Physiol 593.8 and the extent of muscle heterogeneity. Because of this, we carried out calculations of how variation in their assumed values would affect both V O2 and mitochondrial P O2 estimates. In this section, we now use (Fig. 6) the four V MAX values stated in the Methods section that are 4.2, 4.6, 4.9and9.5lmin 1. For mitochondrial P 50 (assumed normal value 0.3 mmhg; Wilson et al. 1977; Gnaiger et al. 1998; Scandurra & Gnaiger, 2010), we also used values of 0.14 and 0.46 mmhg. With heterogeneity of V MAX estimated at 0.1 (logsd V MAX ), we also used values of 0.2 (moderate heterogeneity) and 0.3 (severe heterogeneity). This yielded 4 3 3 or 36 combinations of these variables. It should also be noted that the values chosen reflect substantial relative differences for each variable. Figure 6 shows the results of this analysis in a format where predicted V O2 max is plotted against the corresponding predicted average muscle mitochondrial P O2 so that the effects can be seen for both of these outcome variables. The top panel shows results when V MAX = V O2 max + 10%; the middle panel for V MAX = V O2 max + 20% and the lower panel for V MAX = V O2 max + 30%. In each panel, open squares represent P 50 = 0.14 mmhg, filled circles a P 50 = 0.3 mmhg and open triangles a P 50 = 0.46 mmhg. For any one such P 50,the three connected points in each case reflect, from left to right, logsd V MAX = 0.1 (normal), 0.2 (moderate), and 0.3 (severe heterogeneity). In terms of predicted V O2 max, the consequences of these uncertainties are small. The highest V O2 max in all three panels is 3415 ml min 1, and the lowest is 3192 ml min 1. However, in terms of mean mitochondrial P O2,theeffects are greater. P mo2 is systematically higher for the lowest V MAX (range of P mo2 : 0.8 to 3.7 mmhg, top panel) to lower panel (highest V MAX,rangeofP mo2 : 0.3 to 1.9 mmhg). In summary, the sensitivity analysis displayed in Fig. 6 points out that at any V MAX, increase in mitochondrial P 50 results in systematically higher P mo2 values with almost no effect on the relationship between V O2 max and P mo2.on the other hand, increasing logsd V MAX raises average P mo2 for any set of V MAX and P 50 values. Moreover, the figure clearly indicates that the higher the V MAX,thenarroweris the range of potential mitochondrial P O2 values for a given range of logsd V MAX. Simulated impact of lung and muscle heterogeneity on P mo2 and muscle V O2 in two patients with COPD of different severity: Question 4 As Table 2 shows, lung heterogeneities had been measured in each patient using the multiple inert gas elimination technique (Blanco et al. 2010). The first patient displays features that would be regarded as reflecting moderate COPD, while the second patient represents end-stage COPD and extremely limited exercise capacity. Figure 7 displays the impact of functional lung and muscle heterogeneities on the O 2 pathway in the two COPD patients (moderate severity, left panels and end-stage severity, right panels). The format indicates the values of both predicted V O2 max and P mo2 over a range of values of the same three least certain variables ( V MAX and P 50 of the mitochondrial respiration curve and degree of muscle V MAX heterogeneity, none of which are known for these patients). The less affected patient (left three panels) behaves in a fashion very similar to that shown in Fig. 6 for normal subjects (other than for absolute values of peak V O2 under all conditions). In particular, mitochondrial P O2 values are in the same range as for normal subjects. Variation in P mo2 across muscle regions is considerable (not shown) with a SD of 10 mmhg. For the severely affected patient (right three panels), peak V O2 is insensitive to substantial changes in mitochondrial V MAX, P 50 or the extent of heterogeneity. Only mitochondrial P 50 changeswillhavesomeeffecton mitochondrial P O2, but it is important to note that the scale of mitochondrial P O2 is very different than normal, and is far lower under all conditions, with little variation throughout the muscle (SD of 1 mmhg). Discussion Summary of major findings This study has generated the first integrated model of the O 2 pathway that takes into account all the individual components of O 2 transport and utilization considering both exercising and non-exercising tissues at V O2 max in health and in disease (i.e. COPD), thus allowing for heterogeneity in both lungs and muscle. A graphical user interface to parameterize and simulate the integrative model is freely available at http://sourceforge.net/projects/o2pathway/. The research addresses the separate and combined contributions of functional heterogeneities at pulmonary and at skeletal muscle levels to each of the components of the O 2 pathway. Emphasis is placed on estimating the impact of lung and skeletal muscle V MAX / Q ratio inequalities on overall muscle V O2 and on mitochondrial P O2 at maximal exercise. This was done simulating normal subjects at sea level and altitude, and patients with COPD. The main result from now allowing for lung and muscle heterogeneities is that heterogeneity in muscle affects O 2 availability more than does heterogeneity in the lungs if V MAX is only slightly higher than V O2 max.however,if V MAX is greatly in excess of the potential to deliver O 2, as small muscle mass studies suggest (Andersen & Saltin, 1985), the effects of lung and muscle heterogeneity are similar. On the other hand, the impact of lung heterogeneity on V O2 max is insensitive to the V MAX values used. In addition,

J Physiol 593.8 Effects of V A / Q and V MAX Q heterogeneities on V O2 max 1853 Figure 7. Sensitivity analysis of mitochondrial respiration and muscle heterogeneity estimates in COPD Effects of muscle bioenergetics (mitochondrial respiration V MAX and P 50 parameters) and muscle peripheral heterogeneities on muscle V O2 and mitochondrial P O2 (P mo2 ) for given representatives of mild (left-side panels) and severe (right-side panels) limitation of O 2 transport conditions in chronic obstructive pulmonary disease (COPD). V MAX is conservatively taken as 10% (upper panels), 20% (middle panels) and 30% (lower panels) greater than measured V O2 max in the two COPD patients exercising maximally, selected as representatives of both mild and severe O 2 transport conditions. For each V MAX, three mitochondrial P 50 values (0.14, 0.3 and 0.46 mmhg) and three degrees (0.1, 0.2 and 0.3) of muscle perfusion metabolism inhomogeneity (logsd V MAX ) were evaluated. See text for details.

1854 I. Cano and others J Physiol 593.8 because muscle heterogeneity appears to be less than that in the lung, muscle heterogeneity is predicted to have less of an impact on overall O 2 transport and utilization than observed lung heterogeneity. These findings stress the need for further measurements of skeletal muscle metabolism perfusion heterogeneity in health and disease, and also of assessing parameters of the mitochondrial respiration curve ( V MAX and p 50 ). Implications of functional heterogeneities: lung and skeletal muscle Whether arterial oxygenation falls as a result of altitude in normal subjects or as a result of heterogeneity in the lungs in patients with lung disease (at sea level), it is clear that O 2 availability to the muscles must fall. This has been appreciated for many years, as has the consequence of such reduced O 2 availability for maximal exercise capacity. The other potentially important consequence of reduced O 2 availability is reduced mitochondrial P O2. This may have more than one effect on cellular function. While heterogeneity in the lungs must cause mitochondrial P O2 to fall throughout the muscle (all other factors unchanged), heterogeneity in the muscles will additionally cause mitochondrial P O2 to vary between muscle regions. In those areas with higher than average V MAX in relation to blood flow, mitochondrial P O2 is lower than average, but in regions where V MAX is low in relation to blood flow, mitochondrial P O2 must rise to levels greater than would be seen in the absence of heterogeneity. The consequences of subnormal mitochondrial P O2 are discussed below. There may be corresponding functional consequences of a high mitochondrial P O2 as well. A high P O2 may oppose vasodilatation and restrict local perfusion, thus increasing the V MAX / Q back towards normal and offering an intrinsic mechanism to automatically limit mitochondrial P O2 heterogeneity. A high mitochondrial P O2 will also reduce the capillary-to-mitochondrial P O2 diffusion gradient, reducing diffusive O 2 transport and again working towards restoring the balance between O 2 supply and demand. Furthermore, the high P O2 may exert genomic effects that may be opposite to those expected when P O2 is below normal (see above). The potential therefore exists for regional regulation of gene expression (and thus for adaptive programs) according to local P O2.Finally,when mitochondrial P O2 is elevated, the potential exists for increased ROS generation, just as when P O2 is below normal. If so, muscle regions with both very low and very high V MAX / Q ratios may be at risk of oxidative stress. Biological and clinical implications The multilevel impact of disturbances of cellular oxygenation on cell function is well recognized (Semenza, 2011). Hypoxia- (or hyperoxia-) induced biological alterations may play a significant role on underlying mechanisms in different acute and chronic conditions (Resar et al. 2005). A major conclusion of the present study is that mitochondrial P O2 will vary considerably among muscle regions of different V MAX / Q ratio. When V MAX / Q is low, P mo2 is high, and vice versa. This potential variation in P mo2, which can occur during exercise in healthy individuals, but can also occur on submaximal exercise or even at rest in severe disease states, may have significant biological implications. First, since a low P O2 facilitates local vasodilatation in muscle (Rowell, 1986), those areas with high V MAX in relation to Q, thushavingalowp mo2, may preferentially vasodilate, which means that Q may increase locally and that the V MAX / Q ratio is thus, in part, normalized. In this way, V MAX / Q heterogeneity may be to some extent self-limited. This corresponds to hypoxic pulmonary vasoconstriction in the lungs, where it has been known for many years that a region with a low V A / Q ratio will vasoconstrict, limiting blood flow and tending to push the V A / Q ratio back towards normal. Second, the lower the mitochondrial P O2, the higher will be the P O2 difference between the muscle microvasculature and the mitochondria, other factors being equal. This facilitates the diffusion process, and provides an additional, automatic compensatory mechanism to work in the direction of restoring V O2 when blood flow is low. Whether these self-correcting effects are seen in disease remains to be determined. Potential for estimating P mo2 in health and in disease It may be possible to measure mitochondrial P O2 (P mo2 ) in the future (Mik, 2013) but this promising approach is not yet generally available. If this, or another, method can be developed into a feasible approach for intact humans, it offers the potential for estimating, at least approximately, the parameters of the in vivo mitochondrial O 2 respiration curve. By varying inspired O 2 concentration over as wide a range as safely possible, paired values of V O2 and P mo2 could be measured, and if they were to bracket both the O 2 -dependent and O 2 -independent regions of the mitochondrial respiration curve, there may be enough information to estimate V MAX and P 50. Limitations of the analysis Firstly, as in previous work (Wagner, 1993, 1996; Cano et al. 2013), the entire analysis applies to steady-state conditions (meaning, that O 2 partial pressures are constant in time, as is V O2 itself). Therefore, the analyses should not be extrapolated to transient changes in metabolic rate at the beginning or end of exercise.